Search

Tidal flats are among the ecologically richest areas of the world where sediment composition (e.g. median grain size and silt content) and the macrozoobenthic presence play an important role in the health of the ecosystem. Regular monitoring of environmental and ecological variables is essential for sustainable management of the area. While monitoring based on field sampling is very time-consuming, the predictive performance of these variables using satellite images is low due to the spectral homogeneity over these regions. We tested a novel approach that uses features from a variational autoencoder (VAE) model to enhance the predictive performance of remote sensing images for environmental and ecological variables of tidal flats. The model was trained using the Sentinel-2 spectral bands to reproduce the input images, and during this process, the VAE model represents important information on the tidal flats within its layer structure. The information in the layers of the trained model was extracted to form features with identical spatial coverage to the spectral bands. The features and the spectral bands together form the input to random forest models to predict field observations of the sediment characteristics such as median grain size and silt content, as well as the macrozoobenthic biomass and species richness. The maximum prediction accuracy of feature-based maps was close to 62% for the sediment characteristics and 37% for benthic fauna indices. The encoded features improved the prediction accuracy of the random forest regressor model by 15% points on average in comparison to using just the spectral bands. Our method enhances the predictive performance of remote sensing, in particular the spatiotemporal dynamics in median grain size and silt content of the sediment thereby contributing to better-informed management of coastal ecosystems.

Spaceborne remote sensing performs a significant role today in geological mapping, interpretation and analysis of large areas in a rapid and cost-efficient manner. Spaceborne satellite imagery have been successfully utilized in lithological mapping of arid areas, however, similar approaches for lithological mapping in semi-arid regions with high degree of structural complexity, vegetation and topographic effects are yet to be developed. The primary objective of this study is to combine field-based spectroscopic research with geological remote sensing in obtaining representative lithological maps of the Buëch area in South-East France. To achieve this aim Support Vector Machine (SVM) classifier and Spectral Angle Mapper (SAM) were used on Sentinel-2 and ASTER imagery. Spectroscopic examination of field samples enabled the creation of spectral endmembers representative for the area and detection of local weathering mechanisms. The results derived from SVC classifier showed great pattern-recognition ability and more accurate classification of different lithological groups. SAM derived thematic mapper demonstrated the ability to correctly classify a great portion of blue marl outcrops, by recognising the lack of spectral response displayed by blue marls as being the distinctive indicator of the class. Using and evaluating different satellite imagery shows that Sentinel-2 allows better mapping in this semi-arid area, indicating that the higher spatial resolution of Sentinel-2 has an edge over the higher spectral resolution in the SWIR range of ASTER for this specific setting and methodology. The study has shown that field-based spectroscopic research complements geological remote sensing well. The limiting factor in achieving better classification is the spectral and spatial properties of the satellite imagery. Therefore, future research should consider the integration of hyperspectral data and assimilation of geomorphic products to increase the accuracy of lithological

Landslides are among the most widespread hazards in the word, with special relevance in mountainous regions. In January 1994, a rainfall-triggered rotational landslide mobilised about 28000m3 of material from the hillslopes of the Charonnier basin, in the French department of Hautes-Alpes. Hydrological and geotechnical surveys are classic approaches for stability studies, built upon in situ observation and soil parameterization. Nevertheless, the introduction of remote sensing systems for landslide mapping and monitoring in the last decades yields intersecting results, yet their possibilities are not fully exploited and remains nowadays a promising area of study. Reconstruction of landslide topography and its evolution throughout the years is achievable with the help of photogrammetry techniques. In the present study, historical aerial pictures from 1993 and 1999 are used to generate elevation information of the Charonnier landslide area in raster (DTM) and point cloud (DPC) format. Resolution of the available historical photographs appears to be a crucial factor for the accuracy of the derived products, decreasing their RMSE from 3.4m to 0.14m (900dpi vs. 2400dpi images). With the help of a UAV, which captures on demand, high-resolution images, elevation models for 2016 and 2017 are built with errors around 10cm. 1993, 1999, 2016 and 2017 products are compared to distinguish morphodynamic processes in the short and long term. DTMs corresponding to different dates are subtracted one from the other, to detect areas where terrain elevation has varied. This way, the location of the source area and volume of the displaced mass are drowned. The distance between DPCs for different dates is computed with the M3C2 algorithm. The main outcome of the computation is the delimitation of areas subjected to erosion in the flanks of the Charonnier creek, in the form of gullies, and of riverbed incision of around 10cm/year. A soil sampling campaign and material characterization wer

Primary production in coastal waters is important to monitor due to its implications on water quality, fish stocks and the carbon cycle, amongst others. Due to the economic and political importance of the North Sea, long-term monitoring primary production in this region is essential to fully assess the role of climate change on this process. To apply this to the North Sea, a simple light-dependent primary production model by Cole & Cloern was applied onto in-situ data of water quality parameters and regressed against gross primary production values from OS9, a station off the coast of the Eastern Scheldt. Following calibration at this station and validation of water quality parameters at Rijkswaterstaat stations for the Dutch North Sea, primary production was then successfully modelled for the entire North Sea between 2002-2011 using satellite-derived estimates of water quality parameters (Kd and chlorophyll-a) from Envisat MERIS and surface irradiance (Eo) from a NOAA atmospheric model. Results showed that the various hydrodynamic regions had different primary production values and also experienced variable fluctuations in primary production over the years. It was also found that small changes in the largest hydrodynamic regions had the greatest influence on the North Sea primary production budget. No significant trend was found in the annual primary production budget of the North Sea over the period 2003-2011 (p-value =0.595). The results also showed that modelled primary production values were able to detect shifts in trends as a result of climatic variability, although the period of analysis was quite short to properly assess significant long-terms shifts in primary production trends. Finally, the non-linear response of each region to climatic changes was evident in the annual primary production budget, especially in the case of 2010. A proposed method of conducting future analyses on primary production trends is assessing periods in the North Sea separated by r

Debris-flow fans are widespread and often densely populated landforms that occur in regions with high relief. Avulsions cause a major risk for inhabitants as they occur sporadic and with current knowledge they are relatively unpredictable. Recent studies on debris-flow fan development and avulsion behavior reveal that the magnitude-frequency distribution and the associated flow-magnitude sequence of a debris-flow fan exerts a significant control on the occurrence of avulsions. Additionally, multiple studies also confirm the influence of debris-flow sediment composition on debris-flow morphology. Here, the effect of a change in debris-flow sediment composition per debris flow on the avulsion behavior and debris-flow fan development is investigated. This is done by creating two experimental fans where debris-flow sediment compositions are randomly extracted from a heavy-tailed double-pareto distribution with an average around 25 volumetric percent gravel. The sequence of volumes is distributed along a thin double-pareto distribution, where the majority of flow volumes cluster around a mean value of 5 kilograms sediment weight, but significantly larger and smaller flow volumes still occur sporadically. Comparing the experimental fans created in this study to a fan in De Haas et al. (in review) with similar flow-volume sequence but with uniform composition gives insight into the effect of a change in sediment composition for a debris-flow fan that is simultaneously subjected to a change in volumes. The effect of a sediment composition change on avulsion behavior and debris-flow fan development is best seen on a timescale of a couple of debris-flow events. From this study, three main conclusions can be established. (1) Debris-flow sediment composition is able to enhance or diminish the mobility of a debris flow with a given volume. Here, the volume dictates the main mobility of the debris flow but the mobility is influenced by related debris-flow sediment composition. (2

An approach consisting of different methods is applied to determine the geometry, relevant processes and failure mechanism that resulted in the failure of the Charonnier landslide in 1994. Due to their hazardous nature landslides have been a relevant research topic for decades. Despite that individual landslide events are not as hazardous or catastrophic as for example earthquakes, floods or volcanic eruptions, they occur more frequent and are more widespread (Varnes, 1984). Also in the geological formations of the Terres Noires, in south-east France, it is not necessarily the magnitude of events, rather their frequency of occurrence that makes mass movements hazardous. An extremely wet period, between September 1993 and January 1994, caused a hillslope in the Haute-Alps district to fail. Highly susceptible Terres Noires deposit near Charonnier River failed into a rotational landslide, moving an estimated 107,000 m3 material downslope. Precipitation figures between 1985 and 2015 show a clear pattern of intense rainstorms and huge amounts of precipitation in antecedent rainfall. This suggests that the extreme event on January 6 with 65 mm of rain after the wet months of September, October and December caused the sliding surface to fail. A total of 36 soil samples and 22 saturated conductivity measurements show a decreasing permeability with depth and the presence of macro-pores in the topsoil, supplying lateral flow in extreme rainfall events and infiltration with antecedent rainfall periods. Conventional remote sensing observations (e.g. satellite or aerial), with centimeter resolution, over this relatively vegetated landslide remain challenging due to its relative small size. Therefore, an UAV platform with a compact camera was used to capture the current elevation, structure and geomorphological characteristics at the Charonnier landslide with a 6-centimeter resolution elevation model. Multi View workflow and the Structure from Motion process were used to derive a

Debris-covered glaciers in the Himalayas are relatively unstudied due to the difficulties in fieldwork caused by the inaccessible terrain and the presence of debris layers, which complicate in situ measurements. To overcome these difficulties an unmanned aerial vehicle (UAV) has been deployed multiple times over two debris covered glaciers in the Langtang catchment, located in the Nepalese Himalayas. Using differential GPS measurements and the Structure for Motion algorithm the UAV imagery was processed into accurate high-resolution digital elevation models and orthomosaics for both pre- and post-monsoon periods. These data were successfully used to estimate seasonal surface flow and mass wasting by using cross-correlation feature tracking and DEM differencing techniques. The results reveal large heterogeneity in mass loss and surface flow over the glacier surfaces, which are primarily caused by the presence of surface features such as ice cliffs and supra-glacial lakes. Accordingly, we systematically analyze those features using an object-based approach and relate their characteristics to the observed dynamics. We show that ice cliffs and supra-glacial lakes are contributing to a significant portion of the melt water of debris covered glaciers and we conclude that UAVs have great potential in understanding the key surface processes that remain largely undetected by using satellite remote sensing.

For decades land-use and land-cover (LULC) conversions have had an important impact on land- and ecosystem degradation, accordingly (historical) LULC information is important for the assessment of such impacts. This information can be derived from black-and-white (B&W) aerial photography. Such photography is often visually interpreted, which is a very time-consuming approach. This study shows that machine learning can be applied on only brightness to derive LULC information. Cropland acreage is semi-automatically mapped by means of Geographic Object-Based Image Analysis (GEOBIA) and Random Forest classification in two study sites in Ethiopia and in The Netherlands. The result is a thematic map with two classes: 1) agricultural cropland and 2) other types of land cover. Overall mapping accuracies attained are 90 % and 96 % for the two study areas respectively. This mapping method increases the timeline at which historical cropland expansion can be mapped purely from brightness information in B&W photography up to the 1930s.

The process of rainfall interception by canopies plays an important role in the hydrological cycle. The fraction of rainfall which is intercepted is largely dependent on species-specific vegetation properties, such as leaf- and branch architecture and canopy structure and - density. Because these vegetation properties spatially vary, canopy interception is a spatially variable process. However, these species-specific vegetation properties are often not, or insufficiently included in hydrological models which focus for example on rainfall-runoff processes, soil moisture balances, transpiration and groundwater recharge from rainfall. This could lead to serious modelling uncertainties and - mistakes. This study aimed at improving an existing spatial interception mapping approach by including relations between species-specific vegetation properties and interception. In order to achieve this goal, a two-fold method was applied. First, the influence of species-specific differences of leaf-and canopy structures on the interception process was studied. Fieldwork on branch- and plot-scale was performed in a study region in Vorarlberg (Austria) to find relations between the vegetation properties and interception. Secondly, the ability of a high-resolution remote sensing imagery based model to estimate the spatial variability of canopy interception on regional scale was studied when these species-specific relations were implemented. Promising fieldwork results were gained which describe the relations between vegetation properties and interception per species. Due to these results, it was possible to model spatial variability of canopy interception for the study region in Vorarlberg, thereby taking species-specific vegetation properties into account. Although the results contribute to the improvement of the current approach to quantitatively estimate spatial variability of canopy interception on regional scale, the method needs to be improved and more research is needed in this

It is commonly recognized that large uncertainties exist in modelled biofuel-induced indirect land-use change, but until now, spatially explicit quantification of such uncertainties by means of error propagation modelling has never been performed. In this study, we demonstrate a general methodology to stochastically calculate direct and indirect land-use change (dLUC and iLUC) caused by an increasing demand for biofuels, with an integrated economic - land-use change model. We use the global Computable General Equilibrium model MAGNET, connected to the spatially explicit land-use change model PLUC. We quantify important uncertainties in the modelling chain. Next, dLUC and iLUC projections for Brazil up to 2030 at different spatial scales and the uncertainty herein are assessed. Our results show that cell-based (5 × 5 km2) probabilities of dLUC range from 0 to 0.77, and of iLUC from 0 to 0.43, indicating that it is difficult to project exactly where dLUC and iLUC will occur, with more difficulties for iLUC than for dLUC. At country level, dLUC area can be projected with high certainty, having a coefficient of variation (cv) of only 0.02, while iLUC area is still uncertain, having a cv of 0.72. The latter means that, considering the 95% confidence interval, the iLUC area in Brazil might be 2.4 times as high or as low as the projected mean. Because this confidence interval is so wide that it is likely to straddle any legislation threshold, our opinion is that threshold evaluation for iLUC indicators should not be implemented in legislation. For future studies, we emphasize the need for provision of quantitative uncertainty estimates together with the calculated LUC indicators, to allow users to evaluate the reliability of these indicators and the effects of their uncertainty on the impacts of land-use change, such as greenhouse gas emissions.

An increase of urban flash flood issues has been observed throughout the world in recent years. It has led to enormous losses and people were killed. The increase of urban flash flooding is mainly due to more intensive rain events and more ground surfaces converted into impervious surfaces. Although the runoff mechanism and process have been widely studied, the runoff modeling on the basis of impervious surface mapping is still a new aspect to the study of urban flooding. The main objective of this study is to evaluate various approaches for impervious surface mapping on the basis of satellite images, and evaluating the hydrological impact of impervious surface expansion in Amersfoort by a runoff model. The study area is located at Amersfoort, which is the second largest city in Utrecht Province in the Netherlands. A fast urbanization process took place over the last 30 years. In addition, Amersfoort is a high risk area of urban (flash) flooding because it is located next to an elevated hill in the south-west. The regression modeling method and the Normalized Linear Spectral Mixing Analysis (NLSMA) methods were applied to Landsat images. Accuracy assessment of these two methods showed that impervious areas mapped by NLSMA had a higher accuracy. The Decision Tree Classification (DTC) method was applied to FORMOSAT-2 image. An overall accuracy of 92% was achieved by the error matrix. Uncertainties, advantages and constraints for these three methods were compared and discussed. The runoff responses were evaluated by linking the impervious surfaces maps to a straightforward rainfall-runoff model. A number of one-hour rainfall events with various intensities were worked out and the results showed that the Amersfoort railway station is the most vulnerable place o runoff in case of an intensive rainfall event. The investigated area can bear a one-hour rainfall with a return time of less than two years. Risk maps of buildings for every scenario were worked out. Risk maps ar

Glacial melt water is a source of drinking water for the 1.3 billion people living in downstream area of the Hindu Kush Himalayan region. Himalayan glaciers function as a fresh water buffer as they constantly release water, making it an important resource for people during the dry season. In this research, surface displacements are studied in order to understand more of the melting behaviour of debris-covered glaciers. Understanding these processes will help predict future water levels in rivers. In situ, glaciological measurements are limited due to the inaccessibility of the glacier and need to be interpolated to assess the full surface dynamics. In this study, imagery of the Lirung glacier in Nepal is used that is derived from an unmanned aerial vehicle (UAV). The dataset is ortho-rectified and covers almost a full year including a summer and a winter season. The flight data are May 2013, October 2013 and May 2014. Based on a Structure from Motion process digital elevation models and ortho-rectified mosaic images are created. Some of the glaciers in the Himalaya are covered with debris. This allows tracking the rocks on top of the glacier, where rocks move upon with the ice. The frequency cross-correlation of Cosi-Corr is using the phase in the Fourier domain to track such debris. The output of this correlation is a continuous raster map containing the surface displacement of the glacier. Fine spatial and general seasonal differences can be found in the surface displacement. During the summer, the maximum average displacement is 6 m a-1 while the winter displacement is 2.5 m a-1 at the upper part of the tongue. The lower part is nearly stagnant throughout the year. It is very likely that basal sliding is the dominating process during the higher summer. It is hypothesized that the increment of both temperature and rainfall during the summer causes a lubrication of the bedrock. Other process on the glacier is a yet unknown depression, which can be found to the nort